Age | Commit message (Collapse) | Author | Files | Lines |
|
A netlink dump callback can return a positive number to signal that more
information needs to be dumped or zero to signal that the dump is
complete. In the second case, the core netlink code will append the
NLMSG_DONE message to the skb in order to indicate to user space that
the dump is complete.
The nexthop dump callback always returns a positive number if nexthops
were filled in the provided skb, even if the dump is complete. This
means that a dump will span at least two recvmsg() calls as long as
nexthops are present. In the last recvmsg() call the dump callback will
not fill in any nexthops because the previous call indicated that the
dump should restart from the last dumped nexthop ID plus one.
# ip nexthop add id 1 blackhole
# strace -e sendto,recvmsg -s 5 ip nexthop
sendto(3, [[{nlmsg_len=24, nlmsg_type=RTM_GETNEXTHOP, nlmsg_flags=NLM_F_REQUEST|NLM_F_DUMP, nlmsg_seq=1691394315, nlmsg_pid=0}, {nh_family=AF_UNSPEC, nh_scope=RT_SCOPE_UNIVERSE, nh_protocol=RTPROT_UNSPEC, nh_flags=0}], {nlmsg_len=0, nlmsg_type=0 /* NLMSG_??? */, nlmsg_flags=0, nlmsg_seq=0, nlmsg_pid=0}], 152, 0, NULL, 0) = 152
recvmsg(3, {msg_name={sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, msg_namelen=12, msg_iov=[{iov_base=NULL, iov_len=0}], msg_iovlen=1, msg_controllen=0, msg_flags=MSG_TRUNC}, MSG_PEEK|MSG_TRUNC) = 36
recvmsg(3, {msg_name={sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, msg_namelen=12, msg_iov=[{iov_base=[{nlmsg_len=36, nlmsg_type=RTM_NEWNEXTHOP, nlmsg_flags=NLM_F_MULTI, nlmsg_seq=1691394315, nlmsg_pid=343}, {nh_family=AF_INET, nh_scope=RT_SCOPE_UNIVERSE, nh_protocol=RTPROT_UNSPEC, nh_flags=0}, [[{nla_len=8, nla_type=NHA_ID}, 1], {nla_len=4, nla_type=NHA_BLACKHOLE}]], iov_len=32768}], msg_iovlen=1, msg_controllen=0, msg_flags=0}, 0) = 36
id 1 blackhole
recvmsg(3, {msg_name={sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, msg_namelen=12, msg_iov=[{iov_base=NULL, iov_len=0}], msg_iovlen=1, msg_controllen=0, msg_flags=MSG_TRUNC}, MSG_PEEK|MSG_TRUNC) = 20
recvmsg(3, {msg_name={sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, msg_namelen=12, msg_iov=[{iov_base=[{nlmsg_len=20, nlmsg_type=NLMSG_DONE, nlmsg_flags=NLM_F_MULTI, nlmsg_seq=1691394315, nlmsg_pid=343}, 0], iov_len=32768}], msg_iovlen=1, msg_controllen=0, msg_flags=0}, 0) = 20
+++ exited with 0 +++
This behavior is both inefficient and buggy. If the last nexthop to be
dumped had the maximum ID of 0xffffffff, then the dump will restart from
0 (0xffffffff + 1) and never end:
# ip nexthop add id $((2**32-1)) blackhole
# ip nexthop
id 4294967295 blackhole
id 4294967295 blackhole
[...]
Fix by adjusting the dump callback to return zero when the dump is
complete. After the fix only one recvmsg() call is made and the
NLMSG_DONE message is appended to the RTM_NEWNEXTHOP response:
# ip nexthop add id $((2**32-1)) blackhole
# strace -e sendto,recvmsg -s 5 ip nexthop
sendto(3, [[{nlmsg_len=24, nlmsg_type=RTM_GETNEXTHOP, nlmsg_flags=NLM_F_REQUEST|NLM_F_DUMP, nlmsg_seq=1691394080, nlmsg_pid=0}, {nh_family=AF_UNSPEC, nh_scope=RT_SCOPE_UNIVERSE, nh_protocol=RTPROT_UNSPEC, nh_flags=0}], {nlmsg_len=0, nlmsg_type=0 /* NLMSG_??? */, nlmsg_flags=0, nlmsg_seq=0, nlmsg_pid=0}], 152, 0, NULL, 0) = 152
recvmsg(3, {msg_name={sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, msg_namelen=12, msg_iov=[{iov_base=NULL, iov_len=0}], msg_iovlen=1, msg_controllen=0, msg_flags=MSG_TRUNC}, MSG_PEEK|MSG_TRUNC) = 56
recvmsg(3, {msg_name={sa_family=AF_NETLINK, nl_pid=0, nl_groups=00000000}, msg_namelen=12, msg_iov=[{iov_base=[[{nlmsg_len=36, nlmsg_type=RTM_NEWNEXTHOP, nlmsg_flags=NLM_F_MULTI, nlmsg_seq=1691394080, nlmsg_pid=342}, {nh_family=AF_INET, nh_scope=RT_SCOPE_UNIVERSE, nh_protocol=RTPROT_UNSPEC, nh_flags=0}, [[{nla_len=8, nla_type=NHA_ID}, 4294967295], {nla_len=4, nla_type=NHA_BLACKHOLE}]], [{nlmsg_len=20, nlmsg_type=NLMSG_DONE, nlmsg_flags=NLM_F_MULTI, nlmsg_seq=1691394080, nlmsg_pid=342}, 0]], iov_len=32768}], msg_iovlen=1, msg_controllen=0, msg_flags=0}, 0) = 56
id 4294967295 blackhole
+++ exited with 0 +++
Note that if the NLMSG_DONE message cannot be appended because of size
limitations, then another recvmsg() will be needed, but the core netlink
code will not invoke the dump callback and simply reply with a
NLMSG_DONE message since it knows that the callback previously returned
zero.
Add a test that fails before the fix:
# ./fib_nexthops.sh -t basic
[...]
TEST: Maximum nexthop ID dump [FAIL]
[...]
And passes after it:
# ./fib_nexthops.sh -t basic
[...]
TEST: Maximum nexthop ID dump [ OK ]
[...]
Fixes: ab84be7e54fc ("net: Initial nexthop code")
Reported-by: Petr Machata <[email protected]>
Closes: https://lore.kernel.org/netdev/[email protected]/
Signed-off-by: Ido Schimmel <[email protected]>
Reviewed-by: Petr Machata <[email protected]>
Reviewed-by: David Ahern <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
We're missing empty line between policies.
DPLL will need this.
Tested-by: Vadim Fedorenko <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
When dont-validate flags are filtered out for do/dump op, the list may
be empty. In that case, avoid rendering the validate field.
Fixes: fa8ba3502ade ("ynl-gen-c.py: render netlink policies static for split ops")
Signed-off-by: Jiri Pirko <[email protected]>
Reviewed-by: Jakub Kicinski <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>
|
|
Alderlake N is an E-core only product using Gracemont
micro-architecture. It fits the pre-existing naming scheme perfectly
fine, adhere to it.
Signed-off-by: Peter Zijlstra (Intel) <[email protected]>
Acked-by: Rafael J. Wysocki <[email protected]>
Acked-by: Hans de Goede <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
There are tons of io_uring tests and examples in liburing and on the
Internet. If you're looking for a benchmark, io_uring-bench.c is just an
acutely outdated version of fio/io_uring. And for basic condensed init
template for likes of selftests take a peek at io_uring_zerocopy_tx.c.
Kill tools/io_uring/, it's a burden keeping it here.
Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/7c740701d3b475dcad8c92602a551044f72176b4.1691543666.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>
|
|
This version addresses issues with:
- CPU count display for power domain != 0
- Support more than 8 sockets
- Error on max CPU count exceeds in one request
- Prevent trying CPU 0 hotplug for kernel version 6.5 or later
- Change mem-frequency display to max-mem-frequency
Signed-off-by: Srinivas Pandruvada <[email protected]>
|
|
The mem-frequency displayed by each profile is not the actual memory
frequency of DIMMs, but the maximum the CPU can support.
Change the mem-frequency field to max-mem-frequency.
Signed-off-by: Srinivas Pandruvada <[email protected]>
|
|
get-reg-list test is used to check for KVM registers regressions
during VM migration which happens when destination host kernel
missing registers that the source host kernel has. The blessed
list registers was created by running on v6.5-rc3
Signed-off-by: Haibo Xu <[email protected]>
Reviewed-by: Andrew Jones <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
Add new skips_set members to vcpu_reg_sublist so as to skip
set operation on some registers.
Suggested-by: Andrew Jones <[email protected]>
Signed-off-by: Haibo Xu <[email protected]>
Reviewed-by: Andrew Jones <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
Only do the get/set tests on present and blessed registers
since we don't know the capabilities of any new ones.
Suggested-by: Andrew Jones <[email protected]>
Signed-off-by: Haibo Xu <[email protected]>
Reviewed-by: Andrew Jones <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
No functional changes. Just move the finalize_vcpu call back to
run_test and do weak function trick to prepare for the opration
in riscv.
Suggested-by: Andrew Jones <[email protected]>
Signed-off-by: Haibo Xu <[email protected]>
Reviewed-by: Andrew Jones <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
No functional changes. Just move the reject_set check logic to a
function so we can check for a specific errno. This is a preparation
for support reject_set in riscv.
Suggested-by: Andrew Jones <[email protected]>
Signed-off-by: Haibo Xu <[email protected]>
Reviewed-by: Andrew Jones <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
Add some unfortunate #ifdeffery to ensure the common get-reg-list.c
can be compiled and run with other architectures. The next
architecture to support get-reg-list should now only need to provide
$(ARCH_DIR)/get-reg-list.c where arch-specific print_reg() and
vcpu_configs[] get defined.
Signed-off-by: Andrew Jones <[email protected]>
Signed-off-by: Haibo Xu <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
Split the arch-neutral test code out of aarch64/get-reg-list.c into
get-reg-list.c. To do this we invent a new make variable
$(SPLIT_TESTS) which expects common parts to be in the KVM selftests
root and the counterparts to have the same name, but be in
$(ARCH_DIR).
There's still some work to be done to de-aarch64 the common
get-reg-list.c, but we leave that to the next patch to avoid
modifying too much code while moving it.
Signed-off-by: Andrew Jones <[email protected]>
Signed-off-by: Haibo Xu <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
Update [1] to LLVM BPF backend seeks to enable generation of BPF_ST
instruction when CPUv4 is selected. This affects expected log messages
for the following selftests:
- log_fixup/missing_map
- spin_lock/lock_id_mapval_preserve
- spin_lock/lock_id_innermapval_preserve
Expected messages in these tests hard-code instruction numbers for BPF
programs compiled from C. These instruction numbers change when
BPF_ST is allowed because single BPF_ST instruction replaces a pair of
BPF_MOV/BPF_STX instructions, e.g.:
r1 = 42;
*(u32 *)(r10 - 8) = r1; ---> *(u32 *)(r10 - 8) = 42;
This commit updates expected log messages to avoid matching specific
instruction numbers (program position still could be uniquely
identified).
[1] https://reviews.llvm.org/D140804
"[BPF] support for BPF_ST instruction in codegen"
Signed-off-by: Eduard Zingerman <[email protected]>
Acked-by: Yonghong Song <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>
|
|
The file cgroup_tcp_skb.c contains redundant implementations of the similar
functions (create_server_sock_v6(), connect_client_server_v6() and
get_sock_port_v6()) found in network_helpers.c. Let's eliminate these
duplicated functions.
Changes from v1:
- Remove get_sock_port_v6() as well.
v1: https://lore.kernel.org/all/[email protected]/
Signed-off-by: Kui-Feng Lee <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>
|
|
Kernel 6.5 version deprecated CPU 0 hotplug. This will cause all
requests to fail to offline CPU 0. Check version number of kernel
and ignore CPU 0 hotplug request with debug aid to use cgroup
isolation feature for CPU 0.
Signed-off-by: Srinivas Pandruvada <[email protected]>
|
|
There is a limit on number of CPUs in one request. This is set to 256.
Currently tool silently ignores request for count over 256. Give an
error message to indicate this.
Signed-off-by: Srinivas Pandruvada <[email protected]>
|
|
MAX_PACKAGE_COUNT limits the intel-speed-select to systems with 8 sockets or fewer.
On a system with more than 8 sockets intel-speed-select silently ignores everything
beyond the 8th socket, rendering the tool useless for those systems.
Increase MAX_PACKAGE_COUNT to support systems with up to 32 sockets.
Signed-off-by: Frank Ramsay <[email protected]>
Signed-off-by: Srinivas Pandruvada <[email protected]>
|
|
Fix CPU count display for power domain != 0. In the function
punit_id is always 0, so it never incremented cpu count for power
domain id != 0.
Update punit_id after call to update_punit_cpu_info() to what is
actually received from the kernel.
Signed-off-by: Srinivas Pandruvada <[email protected]>
|
|
The iocg can have three throttled metrics: wait, debt, delay. This patch
add missing wait_ms to IocgStat to show the latest wait_ms of iocg.
As we are here, group iocg usage percents "inflt%" and "usage%" together,
and group iocg throttled metrics "wait", "debt" and "delay" together.
Effect after changes:
nvme0n1 RUN per=50.0ms cur_per=177105.713:v1053528.587 busy= +0 vrate=135.00%:270.00% params=ssd_dfl(CQ)
active weight hweight% inflt% usage% wait debt delay
InterfererGroup0 * 100/ 100 54.28/ 9.09 0.34 24.07 0.00 0.00 0.00
interfered * 84/ 1000 45.72/ 90.91 0.48 41.09 0.00 0.00 0.00
Signed-off-by: Chengming Zhou <[email protected]>
Acked-by: Tejun Heo <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>
|
|
The real vrate iocost inuse is not base_vrate, but the atomic vtime_rate.
We need iocost_monitor tool to display this real vrate that iocost use,
to check if the boosted compensated vrate is normal.
Effect after change:
nvme0n1 RUN per=50.0ms cur_per=172116.580:v1040587.433 busy= +0 \
vrate=135.00%:270.00% params=ssd_dfl(CQ)
^
|
this is real vrate inuse
Signed-off-by: Chengming Zhou <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>
|
|
When I use iocost_monitor on nvme0n1, this error shows up:
"Could not find ioc for nvme0n1"
There is no kobj in struct queue in recent kernel, it seems that the commit
2bd85221a625 ("block: untangle request_queue refcounting from sysfs")
move the queue kobj to struct gendisk.
Fix it by using mq_kobj which is at the same level with queue kobj.
Signed-off-by: Chengming Zhou <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>
|
|
Use rseq_unqual_scalar_typeof() rather than typeof() in macros to remove
the volatile qualifier (if there is one in the input argument), thus
generating better assembly code in those scenarios.
Also add extra brackets around the "p" parameter in RSEQ_READ_ONCE(),
RSEQ_WRITE_ONCE(), and rseq_unqual_scalar_typeof() across architectures
to preserve expectations of operator priority. Here is an example that
shows how operator priority may be an issue with missing parentheses:
#define m(p) \
do { \
__typeof__(*p) v = 0; \
} while (0)
void fct(unsigned long long *p1)
{
m(p1 + 1); /* works */
m(1 + p1); /* broken */
}
Signed-off-by: Mathieu Desnoyers <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Shuah Khan <[email protected]>
|
|
The arm64 load-acquire/store-release macros from the Linux kernel rseq
selftests are buggy. Remplace them by a working implementation.
Signed-off-by: Mathieu Desnoyers <[email protected]>
Cc: Catalin Marinas <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Shuah Khan <[email protected]>
|
|
Allow defining variables and perform cast with a typeof which removes
the volatile and const qualifiers.
This prevents declaring a stack variable with a volatile qualifier
within a macro, which would generate sub-optimal assembler.
This is imported from the "librseq" project.
Signed-off-by: Mathieu Desnoyers <[email protected]>
Signed-off-by: Shuah Khan <[email protected]>
|
|
Ensure that the basic percpu ops tests are effectively built against
mm_cid.
Signed-off-by: Mathieu Desnoyers <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Shuah Khan <[email protected]>
|
|
Andi reported (see link below) a regression when printing the
'duration_time' tool event, where it gets printed as "not counted" for
most of the CPUs, fix it by skipping zero counts for tool events.
Reported-by: Andi Kleen <[email protected]>
Signed-off-by: Ian Rogers <[email protected]>
Tested-by: Andi Kleen <[email protected]>
Tested-by: Arnaldo Carvalho de Melo <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Athira Rajeev <[email protected]>
Cc: Claire Jensen <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Kan Liang <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: https://lore.kernel.org/all/ZMlrzcVrVi1lTDmn@tassilo/
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
core_reg_fixup() complicates sharing the get-reg-list test with
other architectures. Rather than work at keeping it, with plenty
of #ifdeffery, just delete it, as it's unlikely to test a kernel
based on anything older than v5.2 with the get-reg-list test,
which is a test meant to check for regressions in new kernels.
(And, an older version of the test can still be used for older
kernels if necessary.)
Signed-off-by: Andrew Jones <[email protected]>
Signed-off-by: Haibo Xu <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
Rename vcpu_config to vcpu_reg_list to be more specific and add
it to kvm_util.h. While it may not get used outside get-reg-list
tests, exporting it doesn't hurt, as long as it has a unique enough
name. This is a step in the direction of sharing most of the get-
reg-list test code between architectures.
Signed-off-by: Andrew Jones <[email protected]>
Signed-off-by: Haibo Xu <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
print_reg() and its helpers only use the vcpu_config pointer for
config_name(). So just pass the config name in instead, which is used
as a prefix in asserts. print_reg() can now be compiled independently
of config_name().
Signed-off-by: Andrew Jones <[email protected]>
Signed-off-by: Haibo Xu <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
The check doesn't prove much anyway, as the reg lists could be
messed up too. Just drop the check to simplify making print_reg
more independent.
Signed-off-by: Andrew Jones <[email protected]>
Signed-off-by: Haibo Xu <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
The original author of aarch64/get-reg-list.c (me) was wearing
tunnel vision goggles when implementing str_with_index(). There's
no reason to have such a special case string function. Instead,
take inspiration from glib and implement strdup_printf. The
implementation builds on vasprintf() which requires _GNU_SOURCE,
but we require _GNU_SOURCE in most files already.
Signed-off-by: Andrew Jones <[email protected]>
Signed-off-by: Haibo Xu <[email protected]>
Signed-off-by: Anup Patel <[email protected]>
|
|
Commit 3fd7a168bf51 ("perf script: Add 'cgroup' field for output")
added support for printing cgroup path in perf script output.
It was okay if you didn't want any stacks:
$ sudo perf script --comms jpegtran:23f4bf -F comm,tid,cpu,time,cgroup
jpegtran:23f4bf 3321915 [013] 404718.587488: /idle.slice/polish.service
jpegtran:23f4bf 3321915 [031] 404718.592073: /idle.slice/polish.service
With stacks it gets messier as cgroup is printed after the stack:
$ perf script --comms jpegtran:23f4bf -F comm,tid,cpu,time,cgroup,ip,sym
jpegtran:23f4bf 3321915 [013] 404718.587488:
5c554 compress_output
570d9 jpeg_finish_compress
3476e jpegtran_main
330ee jpegtran::main
326e2 core::ops::function::FnOnce::call_once (inlined)
326e2 std::sys_common::backtrace::__rust_begin_short_backtrace
/idle.slice/polish.service
jpegtran:23f4bf 3321915 [031] 404718.592073:
8474d jsimd_encode_mcu_AC_first_prepare_sse2.PADDING
55af68e62fff [unknown]
/idle.slice/polish.service
Let's instead print cgroup on the same line as comm:
$ perf script --comms jpegtran:23f4bf -F comm,tid,cpu,time,cgroup,ip,sym
jpegtran:23f4bf 3321915 [013] 404718.587488: /idle.slice/polish.service
5c554 compress_output
570d9 jpeg_finish_compress
3476e jpegtran_main
330ee jpegtran::main
326e2 core::ops::function::FnOnce::call_once (inlined)
326e2 std::sys_common::backtrace::__rust_begin_short_backtrace
jpegtran:23f4bf 3321915 [031] 404718.592073: /idle.slice/polish.service
8474d jsimd_encode_mcu_AC_first_prepare_sse2.PADDING
55af68e62fff [unknown]
Fixes: 3fd7a168bf514979 ("perf script: Add 'cgroup' field for output")
Signed-off-by: Ivan Babrou <[email protected]>
Acked-by: Ian Rogers <[email protected]>
Acked-by: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: [email protected]
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
To pick up the changes from these csets:
522b1d69219d8f08 ("x86/cpu/amd: Add a Zenbleed fix")
That cause no changes to tooling:
$ tools/perf/trace/beauty/tracepoints/x86_msr.sh > before
$ cp arch/x86/include/asm/msr-index.h tools/arch/x86/include/asm/msr-index.h
$ tools/perf/trace/beauty/tracepoints/x86_msr.sh > after
$ diff -u before after
$
Just silences this perf build warning:
Warning: Kernel ABI header differences:
diff -u tools/arch/x86/include/asm/msr-index.h arch/x86/include/asm/msr-index.h
Cc: Adrian Hunter <[email protected]>
Cc: Borislav Petkov (AMD) <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
This reverts commit 46d21ec067490ab9cdcc89b9de5aae28786a8b8e.
The tests were made with a specific workload, further tests on a
recently updated fedora 38 system with a system wide perf.data file
shows 'perf report' taking excessive time resolving inlines in vmlinux,
so lets revert this until a full investigation and improvement on the
addr2line support code is made.
Reported-by: Jesper Dangaard Brouer <[email protected]>
Acked-by: Artem Savkov <[email protected]>
Tested-by: Jesper Dangaard Brouer <[email protected]>
Cc: Andrii Nakryiko <[email protected]>
Cc: Namhyung Kim <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Alexander Shishkin <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Masami Hiramatsu <[email protected]>
Cc: Milian Wolff <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
An unbound workqueue can be served by multiple worker_pools to improve
locality. The segmentation is achieved by grouping CPUs into pods. By
default, the cache boundaries according to cpus_share_cache() define the
CPUs are grouped. Let's a workqueue is allowed to run on all CPUs and the
system has two L3 caches. The workqueue would be mapped to two worker_pools
each serving one L3 cache domains.
While this improves locality, because the pod boundaries are strict, it
limits the total bandwidth a given issuer can consume. For example, let's
say there is a thread pinned to a CPU issuing enough work items to saturate
the whole machine. With the machine segmented into two pods, no matter how
many work items it issues, it can only use half of the CPUs on the system.
While this limitation has existed for a very long time, it wasn't very
pronounced because the affinity grouping used to be always by NUMA nodes.
With cache boundaries as the default and support for even finer grained
scopes (smt and cpu), it is now an a lot more pressing problem.
This patch implements non-strict affinity scope where the pod boundaries
aren't enforced strictly. Going back to the previous example, the workqueue
would still be mapped to two worker_pools; however, the affinity enforcement
would be soft. The workers in both pools would have their cpus_allowed set
to the whole machine thus allowing the scheduler to migrate them anywhere on
the machine. However, whenever an idle worker is woken up, the workqueue
code asks the scheduler to bring back the task within the pod if the worker
is outside. ie. work items start executing within its affinity scope but can
be migrated outside as the scheduler sees fit. This removes the hard cap on
utilization while maintaining the benefits of affinity scopes.
After the earlier ->__pod_cpumask changes, the implementation is pretty
simple. When non-strict which is the new default:
* pool_allowed_cpus() returns @pool->attrs->cpumask instead of
->__pod_cpumask so that the workers are allowed to run on any CPU that
the associated workqueues allow.
* If the idle worker task's ->wake_cpu is outside the pod, kick_pool() sets
the field to a CPU within the pod.
This would be the first use of task_struct->wake_cpu outside scheduler
proper, so it isn't clear whether this would be acceptable. However, other
methods of migrating tasks are significantly more expensive and are likely
prohibitively so if we want to do this on every work item. This needs
discussion with scheduler folks.
There is also a race window where setting ->wake_cpu wouldn't be effective
as the target task is still on CPU. However, the window is pretty small and
this being a best-effort optimization, it doesn't seem to warrant more
complexity at the moment.
While the non-strict cache affinity scopes seem to be the best option, the
performance picture interacts with the affinity scope and is a bit
complicated to fully discuss in this patch, so the behavior is made easily
selectable through wqattrs and sysfs and the next patch will add
documentation to discuss performance implications.
v2: pool->attrs->affn_strict is set to true for per-cpu worker_pools.
Signed-off-by: Tejun Heo <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Cc: Linus Torvalds <[email protected]>
|
|
Add three more affinity scopes - WQ_AFFN_CPU, SMT and CACHE - and make CACHE
the default. The code changes to actually add the additional scopes are
trivial.
Also add module parameter "workqueue.default_affinity_scope" to override the
default scope and "affinity_scope" sysfs file to configure it per workqueue.
wq_dump.py and documentations are updated accordingly.
This enables significant flexibility in configuring how unbound workqueues
behave. If affinity scope is set to "cpu", it'll behave close to a per-cpu
workqueue. On the other hand, "system" removes all locality boundaries.
Many modern machines have multiple L3 caches often while being mostly
uniform in terms of memory access. Thus, workqueue's previous behavior of
spreading work items in each NUMA node had negative performance implications
from unncessarily crossing L3 boundaries between issue and execution.
However, picking a finer grained affinity scope also has a downside in that
an issuer in one group can't utilize CPUs in other groups.
While dependent on the specifics of workload, there's usually a noticeable
penalty in crossing L3 boundaries, so let's default to CACHE. This issue
will be further addressed and documented with examples in future patches.
Signed-off-by: Tejun Heo <[email protected]>
|
|
configuration
Lack of visibility has always been a pain point for workqueues. While the
recently added wq_monitor.py improved the situation, it's still difficult to
understand what worker pools are active in the system, how workqueues map to
them and why. The lack of visibility into how workqueues are configured is
going to become more noticeable as workqueue improves locality awareness and
provides more mechanisms to customize locality related behaviors.
Now that the basic framework for more flexible locality support is in place,
this is a good time to improve the situation. This patch adds
tools/workqueues/wq_dump.py which prints out the topology configuration,
worker pools and how workqueues are mapped to pools. Read the command's help
message for more details.
Signed-off-by: Tejun Heo <[email protected]>
|
|
Adding get_func_ip test for uprobe inside function that validates
the get_func_ip helper returns correct probe address value.
Tested-by: Alan Maguire <[email protected]>
Signed-off-by: Jiri Olsa <[email protected]>
Acked-by: Yonghong Song <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>
|
|
Adding get_func_ip tests for uprobe on function entry that
validates that bpf_get_func_ip returns proper values from
both uprobe and return uprobe.
Tested-by: Alan Maguire <[email protected]>
Signed-off-by: Jiri Olsa <[email protected]>
Acked-by: Yonghong Song <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>
|
|
Adding support for bpf_get_func_ip helper for uprobe program to return
probed address for both uprobe and return uprobe.
We discussed this in [1] and agreed that uprobe can have special use
of bpf_get_func_ip helper that differs from kprobe.
The kprobe bpf_get_func_ip returns:
- address of the function if probe is attach on function entry
for both kprobe and return kprobe
- 0 if the probe is not attach on function entry
The uprobe bpf_get_func_ip returns:
- address of the probe for both uprobe and return uprobe
The reason for this semantic change is that kernel can't really tell
if the probe user space address is function entry.
The uprobe program is actually kprobe type program attached as uprobe.
One of the consequences of this design is that uprobes do not have its
own set of helpers, but share them with kprobes.
As we need different functionality for bpf_get_func_ip helper for uprobe,
I'm adding the bool value to the bpf_trace_run_ctx, so the helper can
detect that it's executed in uprobe context and call specific code.
The is_uprobe bool is set as true in bpf_prog_run_array_sleepable, which
is currently used only for executing bpf programs in uprobe.
Renaming bpf_prog_run_array_sleepable to bpf_prog_run_array_uprobe
to address that it's only used for uprobes and that it sets the
run_ctx.is_uprobe as suggested by Yafang Shao.
Suggested-by: Andrii Nakryiko <[email protected]>
Tested-by: Alan Maguire <[email protected]>
[1] https://lore.kernel.org/bpf/CAEf4BzZ=xLVkG5eurEuvLU79wAMtwho7ReR+XJAgwhFF4M-7Cg@mail.gmail.com/
Signed-off-by: Jiri Olsa <[email protected]>
Tested-by: Viktor Malik <[email protected]>
Acked-by: Yonghong Song <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull x86/srso fixes from Borislav Petkov:
"Add a mitigation for the speculative RAS (Return Address Stack)
overflow vulnerability on AMD processors.
In short, this is yet another issue where userspace poisons a
microarchitectural structure which can then be used to leak privileged
information through a side channel"
* tag 'x86_bugs_srso' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/srso: Tie SBPB bit setting to microcode patch detection
x86/srso: Add a forgotten NOENDBR annotation
x86/srso: Fix return thunks in generated code
x86/srso: Add IBPB on VMEXIT
x86/srso: Add IBPB
x86/srso: Add SRSO_NO support
x86/srso: Add IBPB_BRTYPE support
x86/srso: Add a Speculative RAS Overflow mitigation
x86/bugs: Increase the x86 bugs vector size to two u32s
|
|
A movsx selftest is added for sign-extension of frame pointer R10.
The verification fails for both privileged and unprivileged
prog runs.
Signed-off-by: Yonghong Song <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Martin KaFai Lau <[email protected]>
|
|
Not used in any other place, so just make it static.
Cc: Adrian Hunter <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Masami Hiramatsu <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/lkml/ZM0pjfOe6R4X%[email protected]/
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
in synthesize_perf_probe_command()
Building perf with EXTRA_CFLAGS="-fsanitize=address" a leak was detected
elsewhere and lead to an audit, where we found that
synthesize_perf_probe_command() may leak synthesize_perf_probe_point()
return on failure, fix it.
Cc: Adrian Hunter <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Masami Hiramatsu <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
to add a probe
Building perf with EXTRA_CFLAGS="-fsanitize=address" a leak is detect
when trying to add a probe to a non-existent function:
# perf probe -x ~/bin/perf dso__neW
Probe point 'dso__neW' not found.
Error: Failed to add events.
=================================================================
==296634==ERROR: LeakSanitizer: detected memory leaks
Direct leak of 128 byte(s) in 1 object(s) allocated from:
#0 0x7f67642ba097 in calloc (/lib64/libasan.so.8+0xba097)
#1 0x7f67641a76f1 in allocate_cfi (/lib64/libdw.so.1+0x3f6f1)
Direct leak of 65 byte(s) in 1 object(s) allocated from:
#0 0x7f67642b95b5 in __interceptor_realloc.part.0 (/lib64/libasan.so.8+0xb95b5)
#1 0x6cac75 in strbuf_grow util/strbuf.c:64
#2 0x6ca934 in strbuf_init util/strbuf.c:25
#3 0x9337d2 in synthesize_perf_probe_point util/probe-event.c:2018
#4 0x92be51 in try_to_find_probe_trace_events util/probe-event.c:964
#5 0x93d5c6 in convert_to_probe_trace_events util/probe-event.c:3512
#6 0x93d6d5 in convert_perf_probe_events util/probe-event.c:3529
#7 0x56f37f in perf_add_probe_events /var/home/acme/git/perf-tools-next/tools/perf/builtin-probe.c:354
#8 0x572fbc in __cmd_probe /var/home/acme/git/perf-tools-next/tools/perf/builtin-probe.c:738
#9 0x5730f2 in cmd_probe /var/home/acme/git/perf-tools-next/tools/perf/builtin-probe.c:766
#10 0x635d81 in run_builtin /var/home/acme/git/perf-tools-next/tools/perf/perf.c:323
#11 0x6362c1 in handle_internal_command /var/home/acme/git/perf-tools-next/tools/perf/perf.c:377
#12 0x63667a in run_argv /var/home/acme/git/perf-tools-next/tools/perf/perf.c:421
#13 0x636b8d in main /var/home/acme/git/perf-tools-next/tools/perf/perf.c:537
#14 0x7f676302950f in __libc_start_call_main (/lib64/libc.so.6+0x2950f)
SUMMARY: AddressSanitizer: 193 byte(s) leaked in 2 allocation(s).
#
synthesize_perf_probe_point() returns a "detachec" strbuf, i.e. a
malloc'ed string that needs to be free'd.
An audit will be performed to find other such cases.
Acked-by: Masami Hiramatsu <[email protected]>
Cc: Adrian Hunter <[email protected]>
Cc: Ian Rogers <[email protected]>
Cc: Jiri Olsa <[email protected]>
Cc: Namhyung Kim <[email protected]>
Link: https://lore.kernel.org/lkml/[email protected]
Signed-off-by: Arnaldo Carvalho de Melo <[email protected]>
|
|
Pull kvm fixes from Paolo Bonzini:
"x86:
- Fix SEV race condition
ARM:
- Fixes for the configuration of SVE/SME traps when hVHE mode is in
use
- Allow use of pKVM on systems with FF-A implementations that are
v1.0 compatible
- Request/release percpu IRQs (arch timer, vGIC maintenance)
correctly when pKVM is in use
- Fix function prototype after __kvm_host_psci_cpu_entry() rename
- Skip to the next instruction when emulating writes to TCR_EL1 on
AmpereOne systems
Selftests:
- Fix missing include"
* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
selftests/rseq: Fix build with undefined __weak
KVM: SEV: remove ghcb variable declarations
KVM: SEV: only access GHCB fields once
KVM: SEV: snapshot the GHCB before accessing it
KVM: arm64: Skip instruction after emulating write to TCR_EL1
KVM: arm64: fix __kvm_host_psci_cpu_entry() prototype
KVM: arm64: Fix resetting SME trap values on reset for (h)VHE
KVM: arm64: Fix resetting SVE trap values on reset for hVHE
KVM: arm64: Use the appropriate feature trap register when activating traps
KVM: arm64: Helper to write to appropriate feature trap register based on mode
KVM: arm64: Disable SME traps for (h)VHE at setup
KVM: arm64: Use the appropriate feature trap register for SVE at EL2 setup
KVM: arm64: Factor out code for checking (h)VHE mode into a macro
KVM: arm64: Rephrase percpu enable/disable tracking in terms of hyp
KVM: arm64: Fix hardware enable/disable flows for pKVM
KVM: arm64: Allow pKVM on v1.0 compatible FF-A implementations
|
|
IRQ_MOVE_CLEANUP_VECTOR is not longer in use. Remove the last traces.
Signed-off-by: Xin Li <[email protected]>
Signed-off-by: Thomas Gleixner <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
|
|
Tune the macros in the using order and align most of them.
Signed-off-by: Zhangjin Wu <[email protected]>
Signed-off-by: Willy Tarreau <[email protected]>
|